by Will McCaffrey, Northrop Grumman

Published by Signal Integrity Journal


The Voltage Regulator Modules (VRMs) are a vital part of any hardware design and critical to system-level power integrity analysis. When available, VRM models provided by the vendors offer a reasonable starting point for a power integrity design analysis, assuming the model properly represents the output impedance. However, these should not be used for design sign-off. The resistor-inductor (R-L) model is typically the most common SPICE representation for the ideal VRM small-signal passive model. Ideal VRM models in SPICE can sometimes provide reasonable first-order approximations for circuit behavior [1]. However, there should be caution in using these models without verification by making measurements. R-L models only include one of six noise sources in a VRM, the output self-impedance.

The input self-impedance, power supply rejection ratio, reverse transfer, input noise current, and output noise current are all important sources of noise that are missed in an R-L model. [2] Furthermore, the small-signal passive model also gives up the dynamic impedance of the VRM due to changes in the load step current.

In this paper, the Analog Devices Inc. (ADI) LTM4650A micro-module with an internal inductor [3] is examined, comparing the generated small-signal passive model from LTpowerCAD [4] to a measured model using the LTM4650A evaluation board. This evaluation board, the DC2603A-A, contains a single LTM4650A with either a single phase or dual phase output configuration. In this measurement, the LTM4650A has two separate outputs (single-phase), with the measured output of one phase set to 1.5V. The LTpowerCAD schematic of the evaluation board is shown in Figure 1.

Click here to read the entire article.